SentenceTransformer based on allenai/specter2_aug2023refresh_base

This is a sentence-transformers model finetuned from allenai/specter2_aug2023refresh_base. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

  • Model Type: Sentence Transformer
  • Base model: allenai/specter2_aug2023refresh_base
  • Maximum Sequence Length: 512 tokens
  • Output Dimensionality: 768 dimensions
  • Similarity Function: Cosine Similarity

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("m7n/discipline-tuned_specter_2_010")
# Run inference
sentences = [
    'The aim of the study is to describe our experience with ultrasound guided drainage of tubo-ovarian abscess with concomitant use of antibiotics in a second level center. Seven women diagnosed with a tubo-ovarian abscess and treated with transvaginal ultrasound guided drainage with concomitant use of antibiotics, between January and January , were reviewed. Intravenous antibiotics were administered as soon as the diagnosis was reached and transvaginal ultrasound guided aspiration of the abscess material was performed within hours with no need of anaesthesia. Transvaginal route was used since it provides a better visualization and access to the region of interest than other ultrasound routes. All cases but one ( %) improved clinically within hours of aspiration and only one required surgery due to refilling of a bilateral tubo-ovarian abscess hours after drainage. Mean hospital stay was days (range - ). No procedure related complications were diagnosed. A follow up ultrasound six months after the drainage showed in cases sonographic markers of chronic tubal inflammatory disease but in all cases the patients remained asymptomatic. Transvaginal ultrasound-guided drainage with concomitant antibiotics appears to be a safe, efficacious and well tolerated procedure in the treatment approach of tubo-ovarian abscess as reported in the literature. We consider this approach as a feasible alternative to surgical drainage whenever indicated.',
    'To compare the usefulness and accuracy of sonographically guided endometrial biopsies. After obtaining informed consents endometrial biopsies were performed using ultrasound guidance in patients followed by operative hysteroscopy. Diagnostic accuracy and treatment efficiency for sono guidance were established. The hysteroscopic procedure was in all cases started by using a fore-oblique mm hysteroscope (Karl Storz®️ Endoscopy-America, Inc. Corporate Pointe, Culver City, CA - ) with a diagnostic sleeve through the cervical os (Karl Storz®️ Endoscopy-America, Inc. Corporate Pointe, Culver City, CA - ), without prior dilatation of the cervix. The catheter used for the polypectomy was an "Intrauterine Access Balloon Catheter" (Cook OB/GYN®️ West Morgan Street, P.O. Box , Spencer, Indiana ). Successful sonographic management of the endometrial pathology had been achieved in patients ( %). Endometrial polyps had been completely removed under sonographic guidance in patients, partially in as confirmed by hysteroscopy. All incompletely removed polyps were of large size (> cm), the remnants were taken out hysteroscopically. Targeted endometrial biopsy was performed under sono guidance in patients. The completion of the procedure was confirmed by hysteroscopy. Targeted endometrial biopsies and polyp removal can be successfully performed under sonographic guidance. Large size endometrial polyps may require hysteroscopy.',
    'The article is devoted to the peculiarities of the paid domestic labor market in the Russian economy. It is shown that this market is characterized by the following features: weak state regulation; a high proportion of internal and external migrants; a wide spread of the shadow economy and informal labor relations; gender differences; the presence in the market of an "elite" segment of workers providing higher-quality and highly paid services, and a segment of workers performing temporary, episodic work. It is proved on the basis of market analysis that there is a predominant demand for skilled labor, and wages are at or above the national average. It is concluded that further efforts are needed to legalize the work of domestic workers within the framework of the state employment policy.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Triplet

  • Datasets: specter_2_ and discipline-tuned_specter_2_010
  • Evaluated with TripletEvaluator
Metric specter_2_ discipline-tuned_specter_2_010
cosine_accuracy 0.9341 0.9357

Training Details

Training Dataset

Unnamed Dataset

  • Size: 40,000 training samples
  • Columns: anchor, positive, and negative
  • Approximate statistics based on the first 1000 samples:
    anchor positive negative
    type string string string
    details
    • min: 75 tokens
    • mean: 231.88 tokens
    • max: 512 tokens
    • min: 86 tokens
    • mean: 228.45 tokens
    • max: 512 tokens
    • min: 83 tokens
    • mean: 238.29 tokens
    • max: 512 tokens
  • Samples:
    anchor positive negative
    Self-report checklists are used to assess computer workstation set up, typically by workers not trained in ergonomic assessment or checklist interpretation.Though many checklists exist, few have been evaluated for reliability and validity.This study examined reliability and validity of the Computer Workstation Checklist (CWC) to identify mismatches between workers' self-reported workstation problems.The CWC was completed at baseline and at month to establish reliability. Validity was determined with CWC baseline data compared to an onsite workstation evaluation conducted by an expert in computer workstation assessment.Reliability ranged from fair to near perfect (prevalence-adjusted bias-adjusted kappa, - ); items with the strongest agreement were related to the input device, monitor, computer table, and document holder. The CWC had greater specificity ( of items) than sensitivity ( of items). The positive predictive value was greater than the negative predictive value for all question... The support of good management is fundamental to the success of any safety and health program. Residential construction is a high-risk industry requiring significant commitment by management to impact day-to-day safety and health challenges. Investigators have evaluated management practices and spending trends in a cohort of residential homebuilders in the Denver metro area of Colorado. Findings suggest that companies significantly increased dollars allocated to support safety and health practices between and . In addition, the HomeSafe Pilot Program has positively impacted financial commitments of partner companies. Resource allocations were significantly greater for specific expense categories when comparing pre to post HomeSafe intervention. This paper presents data on the use of written safety and health programs, safety committees, and workers compensation premium cost containment certification, as well as allocations to safety incentive programs (SIP), personal protective equipme... Abstract Background Traumatic brain injury (TBI) occurs in as many as million people worldwide each year and often results in one or more post-traumatic syndromes, including depression, cognitive, emotional, and behavioral deficits. TBI can also increase seizure susceptibility, as well as increase the incidence of epilepsy, a phenomenon known as post-traumatic epilepsy (PTE). Injury type and severity appear to partially predict PTE susceptibility. However, a complete mechanistic understanding of risk factors for PTE is incomplete. Main body From the earliest days of modern neuroscience, to the present day, accumulating evidence supports a significant role for neuroinflammation in the post-traumatic epileptogenic progression. Notably, substantial evidence indicates a role for astrocytes, microglia, chemokines, and cytokines in PTE progression. Although each of these mechanistic components is discussed in separate sections, it is highly likely that it is the totality of cellular and neur...
    Using a rabbit in vivo joint injury model, the primary objective of the study was to determine if a relationship exists between earlier time to initiation of ketotifen fumarate (KF) treatment and posttraumatic joint contracture (PTJC) reduction. The secondary objective was to determine if a coagulation response could be detected with serial thrombelastography (TEG) analysis following acute trauma in this model.PTJC of the knee were created in skeletally mature, New Zealand White rabbits. Five groups of animals were studied: a control group that received twice daily subcutaneous injections of normal saline and treatment groups that received twice daily subcutaneous injections of KF ( mg/kg) starting immediately, -, -, and -weeks post-injury. After weeks of immobilization, flexion contractures were measured biomechanically. Serial TEG analysis was performed on the control group animals pre-injury and weekly post-injury.The average joint contracture in the Control Group ( ) was higher tha... To compare inpatient compliance with venous thromboembolism prophylaxis regimens.A secondary analysis of patients enrolled in the ADAPT (A Different Approach to Preventing Thrombosis) randomized controlled trial.Level I trauma center.Patients with operative extremity or any pelvic or acetabular fracture requiring venous thromboembolism prophylaxis.We compared patients randomized to receive either low molecular weight heparin (LMWH) mg or aspirin mg BID during their inpatient admission.The primary outcome measure was the number of doses missed compared with prescribed number of doses.A total of patients were randomized to receive either LMWH mg BID ( patients) or aspirin mg BID ( patients). No differences observed in percentage of patients who missed a dose (aspirin: % vs LMWH: %, P = ) or mean number of missed doses ( vs doses, P = ). The majority of patients ( %, n = ) did not miss any doses. Missed doses were often associated with an operation.These data should reassure clinicians th... In treatment of dementia, further to the use of medicine, methodological approaches have shown positive results as to the improvement of the people's condition, by employing cognitive, relational, behavioral stimulation techniques, or intervention on the surroundings. The aim of this research file is to verify the efficacy of BAPNE method as a cognitive and relational stimulation tool, on elderly patients diagnosed with Alzheimer's disease or with other kind of mild to moderate dementia. Scientific research has already given evidence of positive results of the BAPNE method on people with mild impairment, in particular concerning the executive functions. In this experiment, a sample group of elderly patients will undergo a cycle of sessions; the estimation of the quantitative results will be determined by comparing the data of the experimental sample group ( elderly patients), with those of the control group ( elderly patients). The cognitive functions and the executive functions will b...
    Objective To examine the validity and usefulness of pandemic simulations aimed at informing practical decision-making in public health.Methods We recruited a multidisciplinary group of nine experts to assess a case-study simulation of influenza transmission in a Swedish county.We used a non-statistical nominal group technique to generate evaluations of the plausibility, formal validity (verification) and predictive validity of the simulation.A health-effect assessment structure was used as a framework for data collection.Findings The unpredictability of social order during disasters was not adequately addressed by simulation methods; even minor disruptions of the social order may invalidate key infrastructural assumptions underpinning current pandemic simulation models.Further, a direct relationship between model flexibility and computation time was noted.Consequently, simulation methods cannot, in practice, support integrated modifications of microbiological, epidemiological and spati... With the onset of the coronavirus disease (COVID- ) pandemic, public health measures such as physical distancing were recommended to reduce transmission of the virus causing the disease. However, the same approach in all areas, regardless of context, may lead to measures being of limited effectiveness and having unforeseen negative consequences, such as loss of livelihoods and food insecurity. A prerequisite to planning and implementing effective, context-appropriate measures to slow community transmission is an understanding of any constraints, such as the locations where physical distancing would not be possible. Focusing on sub-Saharan Africa, we outline and discuss challenges that are faced by residents of urban informal settlements in the ongoing COVID- pandemic. We describe how new geospatial data sets can be integrated to provide more detailed information about local constraints on physical distancing and can inform planning of alternative ways to reduce transmission of COVID- b... Since , the Australian Aboriginal and Torres Strait Islander Health Performance Framework (HPF) reports have provided information about Indigenous Australians' health outcomes. The HPF was designed, in consultation with Indigenous stakeholder groups, to promote accountability and inform policy and research. This paper explores bridging the HPF as a theoretical construct and the publicly available data provided against its measures. A whole-of-framework, whole-of-system monitoring perspective was taken to summarise eligible indicators at the state/territory level, organised by the HPF's tier and group hierarchy. Data accompanying the and reports were used to compute improvement over time. Unit change and confidence indicators were developed to create an abstract but interpretable improvement score suitable for aggregation and visualisation at scale. The result is an exploratory methodology that summarises changes over time. An example dashboard visualisation is presented. The use of sec...
  • Loss: TripletLoss with these parameters:
    {
        "distance_metric": "TripletDistanceMetric.COSINE",
        "triplet_margin": 0.3
    }
    

Evaluation Dataset

Unnamed Dataset

  • Size: 2,000 evaluation samples
  • Columns: anchor, positive, and negative
  • Approximate statistics based on the first 1000 samples:
    anchor positive negative
    type string string string
    details
    • min: 80 tokens
    • mean: 231.73 tokens
    • max: 509 tokens
    • min: 84 tokens
    • mean: 236.04 tokens
    • max: 512 tokens
    • min: 86 tokens
    • mean: 233.46 tokens
    • max: 512 tokens
  • Samples:
    anchor positive negative
    Abstract Objective This prospective 0year longitudinal study examined the use of coping styles of fathers and mothers of pediatric cancer patients over time and the prospective effects of coping on distress. Methods Psychological distress (General Health Questionnaire) and the use of seven coping styles (Utrecht Coping List: active problem focussing, palliative and passive reaction patterns, avoidance, social support seeking, expression of emotions, and comforting cognition) were assessed in parents shortly after diagnosis, and months, and years later. Results At diagnosis, parents' use of coping styles did not differ from the norm population except more frequent use of support seeking. No significant change over time was found in a palliative reaction pattern. Support seeking declined and emotional expression increased linearly, whereas use of the remaining coping styles decreased, followed by an increase. At years, parents' use differed from the norm population only in less use of ex... Abstract Objective Event centrality, the degree to which a traumatic event is perceived as central to one's identity, has been associated with posttraumatic stress (PTS) symptoms and posttraumatic growth (PTG) outcomes in various trauma samples. Trauma frameworks are widely used to understand the psychological impact of pediatric cancer; however, event centrality has not been studied in this population. We investigated event centrality in pediatric cancer survivors and healthy comparisons, and its relation with PTS and PTG outcomes. Method Cancer survivors, age ( N = ) and healthy comparisons ( N = ) completed the Centrality of Events Scale and PTS and PTG measures in reference to their most traumatic life event. Cancer survivors who first identified a noncancerrelated event repeated all measures in reference to cancer. Results Centrality scores were significantly higher when referencing cancer compared to noncancer events, even in survivors for whom cancer was not rated as most stress... Abstract Introduction To assess the reliability of short versions of the Australian National University Alzheimer's Disease Risk Index (ANUADRI). Methods A short form of the ANUADRI (ANUADRISF) was developed by assessing risk and protective factors with single questions where possible and with short forms of subquestionnaires where available. The tick box form of the ANUADRI (ANUADRITB) was developed with unique questions for each risk and protective factor for Alzheimer's disease. The short versions were evaluated in an independent community sample of participants with a mean age of (SD = , range = ). Results The short versions demonstrated high reliabilities when compared with the ANUADRI. However, the proportion of misclassification was high for some risk factors and particularly for the ANUADRITB. Discussion The ANUADRISF may be considered if less reliable questions from the ANUADRISF can be replaced with more reliable questions from the ANUADRI for risk/protective factors with hig...
    The effects of glucocorticoids on estrogen-induced changes in LH secretion in the ovariectomized rat and on the estrous cycle and gonadotropin levels in the intact female rat were studied. Preliminary experiments showed that multiple injections of dexamethasone or triamcinolone acetonide (TA) inhibited the estradiol benzoate (EB)-induced elevation of LH in the ovariectomized rat. In subsequent experiments, a single injection of TA was found to inhibit the EB-induced elevation in LH in a dose-dependent manner (minimal effective dose, g) when given h after EB but not at times before EB. Single injections of dexamethasone, cortisol, or progesterone given at this time did not alter LH release. TA given h after EB also blocked the estrogen-dependent increase in pituitary responsiveness to LHRH and the priming effect of multiple injections of LHRH. The pituitary response in oil controls given TA was not altered. Cortisol implants which maintained continuously elevated levels of plasma cortis... Abstract Hindbrain adrenergic/noradrenergic nuclei facilitate endocrine and autonomic responses to physical and psychological challenges. Neurons that synthesize adrenaline and noradrenaline target hypothalamic structures to modulate endocrine responses while descending spinal projections regulate sympathetic function. Furthermore, these neurons respond to diverse stress-related metabolic, autonomic, and psychosocial challenges. Accordingly, adrenergic and noradrenergic nuclei are integrative hubs that promote physiological adaptation to maintain homeostasis. However, the precise mechanisms through which adrenaline- and noradrenaline-synthesizing neurons sense interoceptive and exteroceptive cues to coordinate physiological responses have yet to be fully elucidated. Additionally, the regulatory role of these cells in the context of chronic stress has received limited attention. This mini-review consolidates reports from preclinical rodent studies on the organization and function of bra... Abstract This paper will describe the scope of the Drilling, Completion, and Subsea construction activities and the approach taken by the BP Atlantis Wells Delivery Team in planning and execution. The BP Atlantis Wells Delivery Team recognized early that in order to efficiently execute all of the drilling, completion, subsea construction, and tie back operations to the producing facility, a very disciplined Project Planning and Scheduling approach would be required. A group of dedicated, competent scheduling professionals were assigned to the Drilling and Completion (D&C) Team and proved instrumental to the successful outcome. The D&C scheduling professionals complemented the other professional schedulers strategically selected for each of the project's necessary functional teams and key construction sites. The D&C Team started gaining competency in true project management through development and recruitment as early as three years ( ) prior to the start of development operations. Atla...
    A discharging ear is the most common presenting symptom for ENT conditions. However, some degree of hearing loss is always present. In order to compare the degree of hearing impairment with the size and location of the perforation, we made an effort to conduct this study. The purpose of the study is to ascertain whether, and if so, what, a relationship exists between the location and extent of the tympanic membrane perforation and the severity of hearing loss. In a systematic scoping review of randomized controlled trials, each database was subjected to a unique systematic search approach. Utilizing the methodological approaches specified in the Cochrane Handbook for Systematic Reviewers, a systematic scoping review is conducted after selection criteria, with results reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA). Tympanic membrane anomalies are the root cause of various degrees of conducive deafness. The size of the perforat... Most head and neck cancers are derived from the mucosal epithelium in the oral cavity, pharynx andlarynx and are known collectively as head and neck squamous cell carcinoma (HNSCC). Oral cavity cancers are generally associated with tobacco consumption, alcohol abuse,exposure to environmental pollutants and infection with viral agents, namely HPV and EBV or both, whereaspharynx cancers are increasingly attributed to infection with humanpapillomavirus (HPV), primarilyHPV- . Despiteevidence of histological progression from cellular atypia through various degrees of dysplasia,ultimately leading to invasive HNSCC, most patients are diagnosed with late-stage HNSCC without a clinically evident pre malignant lesion. This article reflects on the capacity of Dante's Comedy, through its words and images, to permeate cultures of different eras. It may be viewed as more than a central element of culture, and as an open work characterised by fluidity and change. This essay, after examining cinematographic and literature examples, attempts to show the Comedy as an important piece of evolving semantic structure, able to resettle in many generations' imagery, perhaps even to mark the genealogy of western representation. If Dante can be understood as a classic suitable to be examined in several worlds and times, his Purgatory may be viewed as a cantica that gives voice and body to typical features of modernity in its current phase. Keywords: Sociologia della letteratura, comunicazione, Purgatorio, modernita, industria culturale
  • Loss: TripletLoss with these parameters:
    {
        "distance_metric": "TripletDistanceMetric.COSINE",
        "triplet_margin": 0.3
    }
    

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • learning_rate: 1e-05
  • weight_decay: 0.01
  • num_train_epochs: 1
  • warmup_ratio: 0.1
  • batch_sampler: no_duplicates

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: steps
  • prediction_loss_only: True
  • per_device_train_batch_size: 8
  • per_device_eval_batch_size: 8
  • per_gpu_train_batch_size: None
  • per_gpu_eval_batch_size: None
  • gradient_accumulation_steps: 1
  • eval_accumulation_steps: None
  • torch_empty_cache_steps: None
  • learning_rate: 1e-05
  • weight_decay: 0.01
  • adam_beta1: 0.9
  • adam_beta2: 0.999
  • adam_epsilon: 1e-08
  • max_grad_norm: 1.0
  • num_train_epochs: 1
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.1
  • warmup_steps: 0
  • log_level: passive
  • log_level_replica: warning
  • log_on_each_node: True
  • logging_nan_inf_filter: True
  • save_safetensors: True
  • save_on_each_node: False
  • save_only_model: False
  • restore_callback_states_from_checkpoint: False
  • no_cuda: False
  • use_cpu: False
  • use_mps_device: False
  • seed: 42
  • data_seed: None
  • jit_mode_eval: False
  • use_ipex: False
  • bf16: False
  • fp16: False
  • fp16_opt_level: O1
  • half_precision_backend: auto
  • bf16_full_eval: False
  • fp16_full_eval: False
  • tf32: None
  • local_rank: 0
  • ddp_backend: None
  • tpu_num_cores: None
  • tpu_metrics_debug: False
  • debug: []
  • dataloader_drop_last: False
  • dataloader_num_workers: 0
  • dataloader_prefetch_factor: None
  • past_index: -1
  • disable_tqdm: False
  • remove_unused_columns: True
  • label_names: None
  • load_best_model_at_end: False
  • ignore_data_skip: False
  • fsdp: []
  • fsdp_min_num_params: 0
  • fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
  • fsdp_transformer_layer_cls_to_wrap: None
  • accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
  • deepspeed: None
  • label_smoothing_factor: 0.0
  • optim: adamw_torch
  • optim_args: None
  • adafactor: False
  • group_by_length: False
  • length_column_name: length
  • ddp_find_unused_parameters: None
  • ddp_bucket_cap_mb: None
  • ddp_broadcast_buffers: False
  • dataloader_pin_memory: True
  • dataloader_persistent_workers: False
  • skip_memory_metrics: True
  • use_legacy_prediction_loop: False
  • push_to_hub: False
  • resume_from_checkpoint: None
  • hub_model_id: None
  • hub_strategy: every_save
  • hub_private_repo: None
  • hub_always_push: False
  • gradient_checkpointing: False
  • gradient_checkpointing_kwargs: None
  • include_inputs_for_metrics: False
  • include_for_metrics: []
  • eval_do_concat_batches: True
  • fp16_backend: auto
  • push_to_hub_model_id: None
  • push_to_hub_organization: None
  • mp_parameters:
  • auto_find_batch_size: False
  • full_determinism: False
  • torchdynamo: None
  • ray_scope: last
  • ddp_timeout: 1800
  • torch_compile: False
  • torch_compile_backend: None
  • torch_compile_mode: None
  • dispatch_batches: None
  • split_batches: None
  • include_tokens_per_second: False
  • include_num_input_tokens_seen: False
  • neftune_noise_alpha: None
  • optim_target_modules: None
  • batch_eval_metrics: False
  • eval_on_start: False
  • use_liger_kernel: False
  • eval_use_gather_object: False
  • average_tokens_across_devices: False
  • prompts: None
  • batch_sampler: no_duplicates
  • multi_dataset_batch_sampler: proportional

Training Logs

Epoch Step Training Loss Validation Loss specter_2__cosine_accuracy discipline-tuned_specter_2_010_cosine_accuracy
0 0 - - 0.8939 -
0.02 100 0.1822 0.1227 0.9083 -
0.04 200 0.0858 0.0739 0.9191 -
0.06 300 0.0697 0.0634 0.9251 -
0.08 400 0.0553 0.0584 0.9284 -
0.1 500 0.0539 0.0552 0.9316 -
0.12 600 0.0599 0.0542 0.9329 -
0.14 700 0.0492 0.0494 0.934 -
0.16 800 0.0552 0.0495 0.9341 -
0.18 900 0.051 - - 0.9357

Framework Versions

  • Python: 3.10.12
  • Sentence Transformers: 3.3.1
  • Transformers: 4.49.0.dev0
  • PyTorch: 2.5.1+cu121
  • Accelerate: 1.2.1
  • Datasets: 3.2.0
  • Tokenizers: 0.21.0

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}
Downloads last month
7
Safetensors
Model size
110M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported third-party Inference Providers, and the model is not deployed on the HF Inference API.

Model tree for m7n/discipline-tuned_specter_2_010

Finetuned
(6)
this model

Evaluation results